[zzylol]

Stacked on #1218; part of #1203, tracked in #1207 (PR 2a-3).

Merge unit. This PR merges into main only together with the rest of #1207's merge unit (steps 1a–6, #1207 "Order"): interim items below are deleted by a later PR in the unit.

Why

#1203 decided one operator per candidate, every intermediate result its own memo group. #1218 switched rules 3 and 8 and added the shared add_candidate_over_groups. This PR switches rules 4 (top-k reuse) and 5 (finer aggregate), which still added two-operator chains.

What

• Rule 4 (topk): each top-k consumer copy is one operator over groups: Project over a Limit(k) group over the shared ranking group. The shared Sort and Limit(k_max) ranking producers are found or added with Stage 0's equality (find_or_add_producer) instead of append_producer's origin-and-parts reuse. The exact-only copies of input groups (exact_input_group) are found or added with find_or_add_group; their exact guarantee is what keeps them apart from the group they copy.
• Rule 5 (finer_aggregate): a finer-aggregate read with a compensating filter is Aggregate over a Filter group over the finer group; without a filter it was already one operator.
• Plans unchanged: no snapshot file changes; Stage 2 candidate counts identical on every snapshot input.

New abstractions not in #1207 or the doc: decisions (2026-11-05)

Per #1207's interface policy, these names are not defined by #1207, #1203 or section 3 of memo-groups-properties-requirements.md. #1207's steps 1a–6 merge into main as one merge unit (#1207, Order), so an interim abstraction may stay in this PR only when a later PR of the unit deletes it; none reaches main. Rows marked Pending approval are not deleted by any PR in the unit and need the user's decision.

Abstraction	What it is for	Decision
exact-only copy of a group (exact_input_group's group)	top-k's exact input	Interim, deleted by 2b (an exactness need is the family or the group, #1203)

How

Both rules call reuse::add_candidate_over_groups (#1218) where they called append_reader, and top-k calls reuse::find_or_add_producer where it called append_producer. No new functions.

Representation / rules / enumeration

• (a) Representation. Unchanged types. Candidates of rules 3, 4, 5 and 8 are one operator over groups; rules 1, 2, 6, 7, 9 and step 10 still add chains.
• (b) Rules. Rules 4 and 5 extend the memo through find_or_add_group. Their matching is unchanged: top-k already checked exactness through groups (exact_input_group recurses into input groups), and the finer-aggregate rule matches groups' logical nodes.
• (c) Enumeration. Unchanged.

Key code interfaces added / modified

None. topk.rs and finer_aggregate.rs change only which crate-internal functions they call.

Usage. Neither rule fires on the KLL example; asap_planner --workload kll.json --run-until stage1_pass2 --format text (and the scan-widening workload) print byte-identical output before and after this PR. Two workloads where they fire, asap_planner --run-until stage1_pass2 --format text, Pass 2 output, before (#1218) vs after:

Top-k: topk by (job) (3, sum_over_time(data[1m])) and topk by (job) (5, sum_over_time(data[1m])), exact, every 10 s (data from the topk-rate snapshot workload):

# before
G3:  Aggregate top_k(k=3)  <- G2
    [1]  TopKReuse  Project ← Limit   inputs=[G7]
G4:  Aggregate top_k(k=5)  <- G2
    [1]  TopKReuse  Project ← Limit   inputs=[G7]
G6:  Sort   [0] TopKReuse  Sort   inputs=[G2]
G7:  Limit  [0] TopKReuse  Limit  inputs=[G6]       # the shared k=5 ranking
# after
G3:  Aggregate top_k(k=3)  <- G2
    [1]  TopKReuse  Project           inputs=[G8]
G4:  Aggregate top_k(k=5)  <- G2
    [1]  TopKReuse  Project           inputs=[G9]
G6:  Sort   [0] TopKReuse  Sort   inputs=[G2]
G7:  Limit  [0] TopKReuse  Limit  inputs=[G6]
G8:  Limit  [0] TopKReuse  Limit  inputs=[G7]       # k=3 of the ranking
G9:  Limit  [0] TopKReuse  Limit  inputs=[G7]       # k=5 of the ranking

(A partitioned Limit over a Limit is not legal to share when equal (NonASAPOp::is_shareable_when_equal), so each consumer's Limit is its own group.)

Finer aggregate (SQL, table metrics(ts, city, host, value)): SELECT city, host, SUM(value) FROM metrics WHERE ts > now() - INTERVAL '5' MINUTE GROUP BY city, host and SELECT SUM(value) FROM metrics WHERE ts > now() - INTERVAL '5' MINUTE AND city = 'NYC', exact:

# before
G6:   Aggregate sum(col=3)  <- G5
    [26]  Rollup  Aggregate sum(col=2) ← Filter   inputs=[G2]
# after
G6:   Aggregate sum(col=3)  <- G5
    [26]  Rollup  Aggregate sum(col=2)            inputs=[G12]
G12:  Filter  <- G2
    [0]   Rollup  Filter  inputs=[G2]             # city = 'NYC' over the per-(city, host) sums

Major algorithm changes

None.

Before this PR

Rules 4 and 5 added consumer chains through append_reader and producer groups through append_producer (reused only by origin and budget parts).

After this PR

Rules 4 and 5 add one operator over groups, with groups found by Stage 0's equality.

Deleted

• topk.rs: its append_producer / append_reader calls and their ? error paths (finding a group cannot fail).
• finer_aggregate.rs: its append_reader call.

append_reader and append_producer stay: rollup (6, 7) still uses them (deleted in the rollup PR).

Plan cost, Stage 2 candidates and planning time

UPDATE_SNAPSHOTS=1 left every snapshot file unchanged. Stage 2 physical candidates and --run-until stage3 wall time per snapshot input, #1218 vs this PR, release builds on one host:

snapshot	Stage 2 candidates, before	after	seconds, before	after
o11y_bench/q02	248	248	0.12	0.12
o11y_bench/q05	498	498	0.24	0.24
o11y_bench/q06	251	251	0.12	0.11
o11y_bench/q11	7	7	0.02	0.02
o11y_bench/q12	2	2	0.01	0.01
o11y_bench/q18	737	737	0.29	0.29
o11y_bench/q19	495	495	0.26	0.26
o11y_bench/q20	4	4	0.01	0.01
o11y_bench/q21	368	368	0.11	0.11
o11y_bench/q22	247	247	0.10	0.10
o11y_bench/q23	247	247	0.10	0.10
o11y_bench/q25	495	495	0.26	0.26
o11y_bench/q26	247	247	0.11	0.11
o11y_bench/q27	4	4	0.01	0.01
sketch/shared-quantiles-1h-mvp	1801	1801	0.89	0.88
sketch/nested-windows-15m-1h-mvp	2281	2281	1.23	1.23
sketch/ratio-of-quantiles-1h-mvp	1802	1802	2.54	2.49
sketch/quantile-single-mvp	201	201	0.09	0.08
sketch/quantile-grouped-mvp	108	108	0.06	0.06
sketch/topk-rate-guarantees-ignored-mvp	87	87	0.06	0.05
sketch/fuzz-issue903-raw-retained	135	135	0.08	0.08
sketch/random-mixed-seed-862-w02-q2-q7-mvp	136	136	0.09	0.09
sketch/random-mixed-seed-862-w01-q3-q8-mvp	226	226	0.11	0.11
sketch/random-mixed-seed-862-long-w01-q3-mvp	72	72	0.09	0.09
total			6.99	6.91

Second review fixes (7e3d360)

• exact_input_group no longer overwrites a found group. find_or_add_group may return an existing group. The exact candidates now replace the own candidate only of a group it added. A found group keeps its candidates and gets the new ones through Memo::add_candidate. No test or example reaches the found case: a temporary panic there did not fire in any test below. So plans are unchanged.
• Follows refactor(stage1): summary readouts and inputs as one operator over groups; Memo::add_candidate; one node equality in Pass 1 (#1203 2a-2) #1218's interface changes: add_candidate_over_groups takes a MemoGroupId owner, and find_or_add_producer has no budget-parts argument.
• The rebased dafb6f4 alone uses refactor(stage1): summary readouts and inputs as one operator over groups; Memo::add_candidate; one node equality in Pass 1 (#1203 2a-2) #1218's old signatures, so it compiles only with 7e3d360.

Notation (formulation → code)

Formulation (#1203, #1201, #1207)	Code name in this PR
memo group $G$	MemoGroup, id MemoGroupId
candidate $v$ added to $G$ (over groups)	add_candidate_over_groups, Memo::add_candidate (doc)
a group of the exact implementations of a group (no formulation term)	exact_input_group

Concepts not in the formulation

• Exact-only copy of a group (exact_input_group): a group whose candidates are only the exact implementations of another group, kept apart by its exact guarantee. In the formulation it would be the original group with $r^{\log}$ of its top-k consumer requiring an exact producer (Logical memo candidate as (one operator over groups, logical properties θ) with a requirement per input; no private nodes #1203: a structural need for exact state is the family or the group itself).
• Producer (find_or_add_producer), accuracy budget parts: see refactor(stage1): summary readouts and inputs as one operator over groups; Memo::add_candidate; one node equality in Pass 1 (#1203 2a-2) #1218 / refactor(stage1): Pass 1 adds a memo group for every intermediate result (#1203 PR 2a-1) #1215.

Second review verification (7e3d360)

• cargo test -p asap-types --lib: 146 passed. cargo test -p asap-logical-asap-optimization --lib: 198 passed.
• cargo test -p asap-integration-tests --test integration -- memo_enumeration reuse_rules operator_sharing sharing_issue_regressions finer_aggregate_reuse planner_layering_example having_shared_aggregate stage1_logical_asap_optimization joint_search_equivalence sketch_rollup rollup_controlled production_rollup_workloads: 154 passed, 23 ignored (pre-existing).
• cargo test -p asap-devtools --lib: 18 passed. --test devtools -- asap_planner:: asap_planner_acceptance:: memo_pricing::: 56 passed, 1 ignored.
• UPDATE_SNAPSHOTS=1 cargo test --release -p asap-devtools --test corpus_snapshots -- --include-ignored: 9 passed, no file changed.
• KLL and scan-widening --run-until stage1_pass2 --format text: byte-identical to c4f1267.
• asap_planner --run-until stage1_pass2 --format text: byte-identical to the pre-review head c4f1267 on 35 workloads besides the two examples: the reviewer's 20 workloads, six rollup / scan-widening / avg workloads, five earlier finer-aggregate, rollup, top-k and scan-KLL examples, and random-mixed-seed-862 workloads 01–04.
• cargo fmt --all; cargo clippy --workspace --all-targets -- -D warnings: clean.

Restack on #1218 (b8402c7)

Rebased on #1218's interface alignment: the doc's Memo::add_candidate replaces push_candidate_if_new (same equality; returns the candidate's CandidateForMemoGroupId), and #1218's CI fix of sketch_rollup::quantiles_rolled_up_keep_the_sketch_bound. Plans unchanged.

• cargo fmt --all -- --check, cargo clippy --workspace --all-targets -- -D warnings: clean.
• cargo test -p asap-logical-asap-optimization --lib: 198 passed.
• cargo test -p asap-integration-tests --test integration -- memo_enumeration reuse_rules operator_sharing sharing_issue_regressions finer_aggregate_reuse planner_layering_example having_shared_aggregate stage1_logical_asap_optimization joint_search_equivalence sketch_rollup rollup_controlled production_rollup_workloads topk: 161 passed, 23 ignored (pre-existing).

Restack (2026-11-05)

On #1215's consumer naming (build_of → build_of) and #1218's one node equality in Pass 1.

• cargo fmt --all -- --check; cargo clippy --workspace --all-targets -- -D warnings: clean; cargo test -p asap-logical-asap-optimization --lib: 197 passed.

Viewer

Viewer tests not updated for the group shape (#1203 decision 6); viewer_contract passes.

Verification

• cargo test -p asap-logical-asap-optimization --lib: 198 passed.
• cargo test -p asap-physical-asap-optimization: 29 passed.
• cargo test -p asap-plan-selection --lib: 142 passed.
• cargo test -p asap-devtools --lib: 18 passed.
• cargo test -p asap-integration-tests --test integration -- memo_enumeration reuse_rules operator_sharing sharing_issue_regressions finer_aggregate_reuse planner_layering_example having_shared_aggregate stage1_logical_asap_optimization joint_search_equivalence sketch_rollup rollup_controlled production_rollup_workloads: 154 passed, 23 ignored (pre-existing).
• cargo test -p asap-devtools --test devtools -- asap_planner:: asap_planner_acceptance:: memo_pricing:: viewer_contract::: 58 passed, 1 ignored.
• UPDATE_SNAPSHOTS=1 cargo test --release -p asap-devtools --test corpus_snapshots -- --include-ignored: 9 passed, no file changed.
• cargo fmt --all; cargo clippy --workspace --all-targets -- -D warnings: clean.
• KLL and scan-widening --run-until stage1_pass2 --format text: byte-identical to refactor(stage1): summary readouts and inputs as one operator over groups; Memo::add_candidate; one node equality in Pass 1 (#1203 2a-2) #1218's head.

Left for 2a-4..k

Rollups (6, 7), scan widening (2), tumbling windows (1), state configurations (9) and step 10, then the deletions listed in #1215 and #1218.

[zzylol]

Closed: replaced by #1207's new order (interfaces first, then one PR per stage), starting with PR 0 #1226. The code on this branch is reused in PRs 1–8; the branch is kept.